Google using search to anticipate flu outbreaks

Stephen Withers
Wednesday, 12 November 2008 04:24

Your IT - Home IT

The flu season's arrived in the northern hemisphere, and Google reckons search patterns can provide early warnings of outbreaks.

Last year, Google started exploring the idea of using patterns in search queries to model things happening in the real world. Disease outbreaks seemed like a useful starting point, and the team determined that in the US, certain flu-related queries had a strong correlation with the actual numbers of cases recorded by the Centers for Disease Control and Prevention.

That would be an interesting though not particularly useful observation if not for one thing. The Google-generated estimates of the number of flu cases are available one to two weeks ahead of the official figures.

If the correlation holds, it could give public health authorities additional time to respond to a major outbreak. The sooner action is taken, the more chance there is of restricting the spread of the disease.

This project was possible because Google was able to analyse "hundreds of billions" of searches over the last five years, and reflect the data mining research interests of Google founders Sergey Brin and Lawrence Page.

From an individual's point of view, Google Flu Trends has little more than curiosity value. By the time it shows flu activity is picking up in a particular state, it might be too late to benefit much from vaccination. So if you do see vaccination as a sensible move, why delay it to a time when stocks may already be running short?

But at least you'd still be ahead of people that waited for the official figures as a prompt to go for a jab.

The development is likely to bring fresh attention to Google's "Don't be evil" mantra. Being able to estimate the number of flu cases is unlikely to do any harm, but what if a similar sort of analysis yielded results that could be exploited to the detriment of others?